Little Bash Problem

[firefexx]

Hi, I'm out to write a script which reads a text file and copies all words into another text file if their length is 8.
But I can't get it work :[

#!/bin/sh

for word  in $(cat list1.txt); do

    if [ $(#word) -eq 8 ]; then
        echo $word >> list2.txt
    fi
    
done

Can somebody tell me what is wrong?

[glennzo]

Replace the () with {}

PHP Code:

if [ ${#word} -eq 8 ] ; then 


This seems to work for me.

[firefexx]

Yes it works. Thanks!

But can you explain me why? The syntax confuses me.

[jpollard]

() is used to designate a subprocess to be executed
{} is used to encapsulate a variable that may be confused with characters outside the {}.

$ is used to signify a substition point - hence $( makes no sense.

Now arithemtic expressions are identified by $((<expression>))
might have worked.

[Gareth Jones]

Originally Posted by jpollard

$ is used to signify a substition point - hence $( makes no sense.

Just to clarify, $(command) does make sense, and captures the output of a command. It's a more modern syntax for `command` (with back-ticks rather than single-quotes, which mean a literal string with no internal substitutions in shell-script). However, variables are expanded with $var or ${var}.

So:

# Literal string with internal substitution (spaces do not split tokens):
$ echo "$PATH"
/usr/lib64/ccache:/usr/local/bin:/usr/bin:/bin:/usr/local/sbin:/usr/sbin:/sbin:/home/gareth/.local/bin:/home/gareth/bin

# Literal string without internal substitution (spaces do not split tokens):
$ echo '$PATH'
$PATH

# Capture command output:
$ echo `ls /`
bin boot dev etc home lib lib64 lost+found media mnt opt proc root run sbin srv sys tmp usr var
$ echo $(ls /)
bin boot dev etc home lib lib64 lost+found media mnt opt proc root run sbin srv sys tmp usr var

# Substitute a variable (spaces split tokens unless the variable is double-quoted):
$ echo $PATH
/usr/lib64/ccache:/usr/local/bin:/usr/bin:/bin:/usr/local/sbin:/usr/sbin:/sbin:/home/gareth/.local/bin:/home/gareth/bin
$ echo ${PATH}
/usr/lib64/ccache:/usr/local/bin:/usr/bin:/bin:/usr/local/sbin:/usr/sbin:/sbin:/home/gareth/.local/bin:/home/gareth/bin

# Do integer maths:
$ echo $((1 + 23))
24
$ echo $[1 + 23]
24

---------- Post added at 08:32 PM ---------- Previous post was at 08:27 PM ----------

I'd also make the following (very minor!) optimizations:

#!/bin/sh

for word  in $(< list1.txt); do

    if [ ${#word} -eq 8 ]; then
        echo $word >> list2.txt
    fi
    
done > list2.txt

The first change uses the bash internal file reader rather than running cat as a separate process.

The second causes list2.txt to only be opened and closed once.

[stevea]

Originally Posted by jpollard

() is used to designate a subprocess to be executed

No - wrong - it's a SUB-SHELL - same shell but new environment - NOT a subprocess.

Also () can be used to modify default precedence, pattern matching in a case statement, in bash function declarations, in array assignment ; e.g: FOO=([0]="foo" [1]="bar" [2]="baz")
Used as <(list) or >(list) to create pipes.

{} is used to encapsulate a variable that may be confused with characters outside the {}.

More accurately ....

${parameter}
The value of parameter is substituted. The braces are required when parameter is a positional parameter with more than one digit, or when parameter is followed by a character which is not to be interpreted as part of its name.

So it delimits parameters where there is the possibility of confusion. A shell variable is one type of parameter, and that's only one of several uses for braces. Brace expansion, function declaration, same-shell lists, ...

Originally Posted by Gareth Jones

Just to clarify, ....

Excellent changes.

[Gareth Jones]

Originally Posted by stevea

No - wrong - it's a SUB-SHELL - same shell but new environment - NOT a subprocess.

A "( command )" sub-shell is generally implemented as a separate child process of the main shell though. "{ command ;}" is similar, but tries to run commands in the current shell whenever possible (pipes for example make that difficult):

[gareth@gareth-desktop ~]$ echo $BASHPID  # Main shell process ID.
19613
[gareth@gareth-desktop ~]$ ( echo $BASHPID )  # Subshell has new PID.
19717
[gareth@gareth-desktop ~]$ { echo $BASHPID; }  # Should have same PID.
19613
[gareth@gareth-desktop ~]$ echo hello | { echo $BASHPID; cat ;}  # Pipe forces sub-shell.
19761
hello

Note that $$ always expands to the main shell's PID, so you have to use $BASHPID to catch the sub-shell.

[RupertPupkin]

I prefer to use a while loop:

#!/bin/bash
while read -r word; do
   if [ ${#word} -eq 8 ]; then
      echo $word
   fi
done < list1.txt > list2.txt

[stevea]

Originally Posted by Gareth Jones

A "( command )" sub-shell is generally implemented as a separate child process of the main shell though. "{ command ;}" is similar, but tries to run commands in the current shell whenever possible (pipes for example make that difficult):

Thanks - I haven't examinined the implementation in many years - but it it's create using ...
strace -f -etrace=clone,execve bash

So both "(cmd)" and "cmd&" so a fork a subprocess currently, and neither does an exec (aside from any exceutable cmd exec). "{cmd;}", as expected executes w/o any extra process creation.


I don't think your comments about pipes is clear (and maybe not accurate). We expect pipes "|" to create new processes .... for example
ls | wc -l
vs
{ ls | wc -l; }

have the same process creation events. Ignoring the execve() calls for the executables, there is always one clone() "aka fork" call per pipe .....

(output line length below trimmed for clarity)

[stevea@crucibulum Desktop]$ ls | wc -l
clone(Process 24114 attached
[pid 23534] clone(Process 24115 attached
Process 23534 suspended
[pid 24114] execve("/bin/ls", ["ls"], [/* 49 vars */]) = 0
[pid 24115] execve("/usr/bin/wc", ["wc", "-l"], [/* 49 vars */]) = 0
17
Process 23534 resumed
Process 24114 detached
Process 24115 detached

[stevea@crucibulum Desktop]$ { ls | wc -l; }
clone(Process 24116 attached
[pid 23534] clone(Process 24117 attached
Process 23534 suspended
[pid 24116] execve("/bin/ls", ["ls"], [/* 49 vars */]) = 0
[pid 24117] execve("/usr/bin/wc", ["wc", "-l"], [/* 49 vars */]) = 0
Process 23534 resumed
Process 24116 detached
17
Process 23534 suspended
Process 23534 resumed
Process 24117 detached

[firefexx]

Ok, thanks for all the answers

[Gareth Jones]

Originally Posted by RupertPupkin

I prefer to use a while loop:

#!/bin/bash
while read -r word; do
   if [ ${#word} -eq 8 ]; then
      echo $word
   fi
done < list1.txt > list2.txt

That's the best way I think.

---------- Post added at 02:55 PM ---------- Previous post was at 02:51 PM ----------

Originally Posted by stevea

I don't think your comments about pipes is clear (and maybe not accurate). We expect pipes "|" to create new processes

Yes, I was probably going into too much detail and then over-simplifying it back. The reason I thought of it is that I remember some other Bourne shell derivatives (ksh maybe?) can run the last part of a pipe-line in the top-level shell so that you can read into shell variables etc. for use later. In Bash you'd have to use process substitution instead I think. But none of that directly relates to sub-shell syntaxes of course.